I Think, Therefore I Am: Why AI Doesn’t Think

 AI Think, Therefore I Am

 

A graduate student of mine is planning a classroom intervention for an applied research course. They want to address off-task behaviors during free-work periods in class. It’s a familiar problem, but one that I think is more difficult to define than it seems. You see, it’s easy to label everything coursework-related as good and as bad anything non-coursework-related. Doing so, however, assumes that what happens in class is a zero-sum game: that anything that isn’t working must be taking away from work. But that’s not how learning happens.

 

Looking back now almost 20 years at my time in college, my favorite course was a lab science I took my senior year. It was one of those “what’s left on my check sheet?” sorts of courses. It was Physics I, and it met at 9:30am, which was difficult for me as a late night restaurant worker. The teacher organized classtime around group work. We were paired in groups of three and four, and we were given the entire class period to complete a worksheet of problems. We spent about 10% of the time answering questions and the other 90% just dicking around.

 

On the surface, it would be easy to say that we wasted 90% of our classtime. "Just think of how much MORE we could have accomplished had we been on task” goes this line of reasoning. But that ignores the social element. We were getting to know each other, and those relationships made attending class and doing the work more tolerable. Enjoyable, really. So it wasn’t time wasted, after all.

 

So this graduate student was describing the problems with and factors relevant to off-task classroom behaviors—how its distracting to other learners and stems from low attention-spans. But the description never went beyond this sort of textbook definition of what an off-task behavior is. The description was AI-heavy, and I found myself disappointed. There was no thought of how the off/on dichotomy is simplistic, or how excellent productivity doesn’t mean 100% productive. (I’ve published eight books and three dozen articles in 12 years, but most of my writing time is spent on so called off-task behaviors. Does that mean I should have published 24 books and 100 articles already?) There was no thought at all, I’d wager. And that’s the problem with AI thinking: it’s not really thinking at all.

 

AI thinking focuses on definitions of concepts, I’d surmise. I imagine that key words are identified in your prompt, and AI scans the web for crowd-sourced definitions. (I’m not sure if it just grabs a popular blog post or if it samples from hundreds on the same topic. Probably the former, but I’m being admittedly pessimistic.) Then the keywords are connected in whatever ways are most common. At least that’s how the AI-influenced student essays always come to me. It’s like a concept map that gets created. Something like this:

 

 

Then a narrative is made in-between the dots. Off-task is defined in a pretty generic way: “Off-task behaviors can influence classtime productivity, efficiency, student motivation, and student achievement, which makes it an important topic for teachers to examine.” Can you see how I never really said anything here?

 

There’s no thinking in this mental map. I quite literally filled out circles with whatever related concepts came to mind, such as “ADHD is connected to mental health.” Now look what happens when I connect those two ideas in AI-speak: “ADHD is a mental health diagnosis, and this can complicate learning—especially on tasks that require focus.” I’m just saying what ADHD is and how what it is might unfold in a classroom.

 

It’s like saying of the skeletal system, “The sternum is a bone that connects the ribs in the center of the chest. It can become damaged, which may lead to any of a number of health problems.” I haven’t done any thinking here, either.

 

By thinking, I mean it the way Martin Heidegger and Max Wertheimer meant it: thinking requires that the thinker be open and receptive to what it is they’re thinking about. For Heidegger, insights present themselves—present like a gift for which we are thankful (thank or think). For Wertheimer, an insight is a recognition of the most integral feature of the problem we’re thinking about—the feature that reveals the essence of the problem. I think when I am open and receptive to something, such as whether and how students are on or off task in the classroom. More specifically, I’m open to the nuances in what it actually means to be on or off task, and even if that’s my question at all.

 

When I think, I almost never stay in the same place. I start off interested in, say, first generation college students. With each thing I learn, my understanding of first gen college students is forever changed. I don’t even mean the same thing anymore when I say FGCS. That’s what happens when I think: there are consequences. I just finished writing a literature review about Overtraining Syndrome in Endurance Athletes. I followed my thinking wherever it took me, which was to a 19^th century medical disorder called neurasthenia. I’m still not sure the extent of the connection, but there’s something there. 

 

For fun, I asked AI to explain the connection between OTS and neurasthenia. What it produced involved zero thinking—just map of definitions. MS CoPilot has promised me that, while the two concepts are similar, “they are not the same thing.” (Emphasis original.) Kind of reveals that AI is incapable of creating a new hypothesis or generating a new insight.

 

To bring this idea back to where I started: it is in the openness to ideas and to possibility—it is in thinking— that I become who I am. Otherwise I’m nothing but a bundle of discrete thoughts that can be connected by lines, but which in their connection the thoughts remain unchanged. AI thinking isn’t thinking; it's just information connected to other information. But humans can think, and in their thinking they are changed. It is something to be thankful for.

 

This teacher will remain unchanged in their research and in their course development if they refuse to think about the problems/solutions they encounter. With AI, they will remain just as they are, as will their students.